Using satellite images and image recognition algorithms, researchers have succeeded in counting the trees distributed south of the Sahara and in the Sahel. A study, detailing the approach, was published Wednesday, October 14 in the journal Nature.
Many trees are present in this very vast area but it is difficult to identify them because, while some form forests, others are more isolated. However, all these trees are essential for the proper functioning of the ecosystem even if scientists have little data on this vegetation. It is these isolated trees with a plant surface of more than 3 m² and located on an area of 1.3 million km² in the western African Sahara, the Sahel and the subhumid zones that the researchers decided to identify.
Satellite images analyzed by an AI
A project impossible to achieve without automation tools. First, the researchers collected images of the area concerned, in particular via the European Copernicus program, which offers free access to data from around 30 sentinel Earth observation satellites. Very high resolution images were also acquired to complete the database.
Then all these images were analyzed using deep learning and image recognition algorithms in order to distinguish the trees in these photos. More than 1.8 billion trees have been detected, or 13.4 trees per hectare with an average plant cover of 12 m², the researchers detail.
These trees, many more numerous than scientists imagined, are essential because they store carbon, provide food and provide refuge for humans and wildlife. Scientists will be able to better follow the evolution of this region and estimate more precisely the rates of deforestation and desertification. The next step will be to precisely map the location and size of all the trees, and why not specify their species.